wechat-cli-local-data

Warn

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: MEDIUMPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill requires administrative execution (sudo on macOS/Linux and elevated terminals on Windows) to initialize the tool and scan process memory for database encryption keys. Additionally, it automates or instructs the forced re-signing of the WeChat application with the com.apple.security.get-task-allow entitlement, lowering the application's native protections against debugging and injection.
  • [EXTERNAL_DOWNLOADS]: The installation guidelines prompt the download of unverified software packages across multiple registries, including the npm package @canghe_ai/wechat-cli, the PyPI package wechat-cli, and a GitHub repository from an unrelated third-party account (freestylefly).
  • [COMMAND_EXECUTION]: The skill relies extensively on executing external CLI binaries (wechat-cli) to access local files, sessions, configurations, and chat databases.
  • [INDIRECT_PROMPT_INJECTION]: The skill reads untrusted external data (WeChat messages, contact bios, group chat logs) and feeds it into the agent's context, exposing the system to potential multi-step or tool-output injection payloads.
  • Ingestion points: Incoming data streams through commands like wechat-cli history, wechat-cli search, and wechat-cli new-messages in SKILL.md.
  • Boundary markers: Absent; no specific delimiters or instruction safety guards are defined to segregate chat text from system instructions.
  • Capability inventory: The tool runs local binary execution workflows via the shell.
  • Sanitization: Absent; no parsing or validation safeguards are defined to strip executable or instructional code blocks from chat transcripts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 15, 2026, 11:03 AM
Security Audit — agent-trust-hub — wechat-cli-local-data